• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2009.04a
• /
• pp.304-307
• /
• 2009

본 논문에서는 3차원 모델링을 이용하여 원공을 갖는 알루미늄 판의 패치 보강효과에 대해 알아보고자 한다. 구조물의 노후화로 인해 높은 응력을 받는 부재의 응력 특이점에서 내구력이 급격하게 저하되거나 때로는 부재의 정적파괴를 유발시키는 원인을 제공한다. 이로 인해 과거에는 손상된 모재에 보강 재료를 연결시키기 위하여 리벳 또는 볼트와 같은 기계적 연결을 통해 보강하였으나 최근에는 접착패치보강 기법이 그 주류를 이루고 있다. 패치 보강시 일면 패치 보강으로 인하여 면외 휨 효과가 발생된다. 판의 두께 방향에 따른 응력집중계수를 별도로 분석하였다. 기존의 3차원 솔리드 요소는 해의 정확성은 뛰어난 반면에 상당한 컴퓨터 시간을 요구하는 단점을 가지고 있다. 이러한 문제를 극복하기 위해서, 본 논문에서는 각 층의 변위장을 2차원 형상함수와 1차원 형상함수의 조합으로 구성하여, 면내거동에 대한 p-세분화와 면외거동에 대한 p-세분화를 분리시키는 방식을 취한다. 또한, 에너지 함수의 적분시 Gauss-Lobatto 적분법을 사용하여 절점의 위치에서의 응력점을 구하는 경우, 외삽과정을 계산하는 단계를 생략하면서도, 해의 정확성 측면에서는 거의 차이가 없기 때문에 좀 더 효율적인 수치적분이 될 수 있다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.6
• /
• pp.2806-2825
• /
• 2018

Unsupervised learning has shown good performance on image, video and audio classification tasks, and much progress has been made so far. It studies how systems can learn to represent particular input patterns in a way that reflects the statistical structure of the overall collection of input patterns. Many promising deep learning systems are commonly trained by the greedy layerwise unsupervised learning manner. The performance of these deep learning architectures benefits from the unsupervised learning ability to disentangling the abstractions and picking out the useful features. However, the existing unsupervised learning algorithms are often difficult to train partly because of the requirement of extensive hyperparameters. The tuning of these hyperparameters is a laborious task that requires expert knowledge, rules of thumb or extensive search. In this paper, we propose a simple and effective unsupervised feature learning algorithm for image classification, which exploits an explicit optimizing way for population and lifetime sparsity. Firstly, a sparse target matrix is built by the competitive rules. Then, the sparse features are optimized by means of minimizing the Euclidean norm ($L_2$) error between the sparse target and the competitive layer outputs. Finally, a classifier is trained using the obtained sparse features. Experimental results show that the proposed method achieves good performance for image classification, and provides discriminative features that generalize well.

• Advanced Composite Materials
• /
• v.18 no.4
• /
• pp.297-314
• /
• 2009

A design method is proposed to optimize the stacking sequence of laminated composite plates for desired vibration characteristics. The objective functions are the natural frequencies of the laminated plates, and three types of optimization problems are studied where the fundamental frequency and the difference of two adjacent frequencies are maximized, and the difference between the target and actual frequencies is minimized. The design variables are a set of discrete values of fiber orientation angles with prescribed increment in the layers of the plates. The four lamination parameters are used to describe the bending property of a symmetrically laminated plate, and are optimized by a gradient method in the first stage. A new technique is introduced in the second stage to convert from the optimum four lamination parameters into the stacking sequence that is composed of the optimum fiber orientation angles of all the layers. Plates are divided into sub-domains composed of the small number of layers and designed sequentially from outer domains. For each domain, the optimum angles are determined by minimizing the errors between the optimum lamination parameters obtained in the first step and the parameters for all possible discrete stacking sequence designs. It is shown in numerical examples that this design method can provide with accurate optimum solutions for the stacking sequence of vibrating composite plates with various boundary conditions.

• Structural Engineering and Mechanics
• /
• v.6 no.5
• /
• pp.497-516
• /
• 1998

A relatively simple two-dimensional multilayered shell model is presented for predicting both global quantities and stress distributions across the thickness of multilayered thick shells, that is based on a third-order zig-zag approach. As for any zig-zag model, the layerwise kinematics is accounted for, with the stress continuity conditions at interfaces met a priori. Moreover, the shell model satisfies the zero transverse shear stress conditions at the upper and lower free surfaces of the shell, irrespective of the lay-up. By changing the parameters in the displacement model, some higher order shell models are obtained as particular cases. Although it potentially has a wide range of validity, application is limited to cylindrical shell panels in cylindrical bending, a lot of solutions of two-dimensional models based on rather different simplyfying assumptions and the exact three-dimensional elasticity solution being available for comparisons for this benchmark problem. The numerical investigation performed by the present shell model and by the shell models derived from it illustrates the effects of transverse shear modeling and the range of applicability of the simplyfying assumptions introduced. The implications of retaining only selected terms depending on the radius-to-thickness ratio are focused by comparing the present solutions to the exact one and to other two-dimensional solutions in literature based on rather different simplyfying assumptions.

• International Journal of Aerospace System Engineering
• /
• v.3 no.1
• /
• pp.10-16
• /
• 2016

A trigonometric Layerwise higher order shear deformation theory (TLHSDT) is developed and implemented for free vibration and buckling analysis of laminated composite and sandwich plates by analytical and finite element formulation. The present model assumes parabolic variation of out-plane stresses through the depth of the plate and also accomplish the zero transverse shear stresses over the surface of the plate. Thus a need of shear correction factor is obviated. The present zigzag model able to meet the transverse shear stress continuity and zigzag form of in-plane displacement continuity at the plate interfaces. Hence, botheration of shear correction coefficient is neglected. In the case of analytical method, the governing differential equation and boundary conditions are obtained from the principle of virtual work. For the finite element formulation, an efficient eight noded $C^0$ continuous isoparametric serendipity element is established and employed to examine the dynamic analysis. Like FSDT, the considered mathematical model possesses similar number of variables and which decides the present models computationally more effective. Several numerical predictions are carried out and results are compared with those of other existing numerical approaches.

• Proceedings of the Korean Society of Propulsion Engineers Conference
• /
• 1997.11a
• /
• pp.29-29
• /
• 1997

각광 받는 구조재료인 섬유강화 복합적층재에 대한 기계적 체결 거동은 본질적인 재료의 이방성에 의해서 파단강도가 파단 모우드와 매우 밀접한 관련을 갖는 것으로 알려져 있다. 따라서, 복합적층판 체결부의 정밀 구조 설계에서는 단순화에 따른 오차를 줄이고 정밀해에 의한 설계 및 해석이 요청된다. 특히, 층간응력 성분을 무시할 수 없는 두께를 갖는 복합적층 판의 기계적 체결부 해석이나 실제 구조물의 체결부에서 발생하는 굽힘이나 비틀림과 같은 하중 상태를 묘사하기 위해서도 정밀한 3차원 응력 해석은 필요하다. 하지만, 지금까지 기계적 체결부의 거동에 관한 연구는 층간응력 성분들을 어느정도 무시할 수 있는 얇은 평판에 대한 2차원 응력해석에 주로 국한되어 왔으며, 일부 수행된 체결부에 대한 3차원 응력 해석의 경우 여러 단점을 갖는 3차원 연속체 요소에 의한 유한요소 해석이 수행되었을 뿐이다.본 연구는 층간응력 성분들을 무시할 수 없는 두께를 갖는 복합적층판의 기계적 체결부 해석에 지금까지 사용되어온 3차원 연속체 요소에 의한 유한요소 방법이 갖는 단점들을 개선한 Layerwise 유한요소법을 이용하여 3차원 응력해석을 수행하였다. 특히, 선형상보성원리에 근거한 최적설계 기법을 응용하여, 기계적 체결시 핀과 적층판의 홀 사이에 발생하는 하중 전달 과정을 모사하고, 접촉력에 의한 홀 주위의 복잡하고 국부적인 응력 집중현상을 규명하여본다.

• Structural Engineering and Mechanics
• /
• v.42 no.4
• /
• pp.471-488
• /
• 2012

An efficient one dimensional finite element model has been presented for the dynamic analysis of composite laminated beams, using the efficient layerwise zigzag theory. To meet the convergence requirements for the weak integral formulation, cubic Hermite interpolation is used for the transverse displacement ($w_0$), and linear interpolation is used for the axial displacement ($u_0$) and shear rotation (${\psi}_0$). Each node of an element has four degrees of freedom. The expressions of variationally consistent inertia, stiffness matrices and the load vector are derived in closed form using exact integration. The formulation is validated by comparing the results with the 2D-FE results for composite symmetric and sandwich beams with various end conditions. The employed finite element model is free of shear locking. The present zigzag finite element results for natural frequencies, mode shapes of cantilever and clamped-clamped beams are obtained with a one-dimensional finite element codes developed in MATLAB. These 1D-FE results for cantilever and clamped beams are compared with the 2D-FE results obtained using ABAQUS to show the accuracy of the developed MATLAB code, for zigzag theory for these boundary conditions. This comparison establishes the accuracy of zigzag finite element analysis for dynamic response under given boundary conditions.

• ETRI Journal
• /
• v.44 no.5
• /
• pp.849-858
• /
• 2022

Dynamic voltage frequency scaling (DVFS) has been widely adopted for runtime power management of various processing units. In the case of neural processing units (NPUs), power management of neural network applications is required to adjust the frequency and voltage every layer to consider the power behavior and performance of each layer. Unfortunately, DVFS is inappropriate for layer-wise run-time power management of NPUs due to the long latency of voltage scaling compared with each layer execution time. Because the frequency scaling is fast enough to keep up with each layer, we propose a layerwise dynamic frequency scaling (DFS) technique for an NPU. Our proposed DFS exploits the highest frequency under the power limit of an NPU for each layer. To determine the highest allowable frequency, we build a power model to predict the power consumption of an NPU based on a real measurement on the fabricated NPU. Our evaluation results show that our proposed DFS improves frame per second (FPS) by 33% and saves energy by 14% on average, compared with DVFS.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.617-621
• /
• 2021

의미역 결정은 문장 속에서 서술어와 그 논항의 관계를 파악하며, '누가, 무엇을, 어떻게, 왜' 등과 같은 의미역 관계를 찾아내는 자연어 처리 기법이다. 최근 수행되고 있는 의미역 결정 연구는 주로 말뭉치를 활용하여 딥러닝 학습을 하는 방식으로 연구가 이루어지고 있다. 최근 구글에서 개발한 사전 훈련된 Bidirectional Encoder Representations from Transformers (BERT) 모델이 다양한 자연어 처리 분야에서 상당히 높은 성능을 보이고 있다. 본 논문에서는 한국어 의미역 결정 성능 향상을 위해 한국어의 언어적 특징을 고려하며 사전 학습된 SNU KR-BERT를 사용하면서 한국어 의미역 결정 모델의 성능을 살펴보였다. 또한, 본 논문에서는 BERT 모델에서 과연 어떤 히든 레이어(hidden layer)에서 한국어 의미역 결정을 더 잘 수행하는지 알아보고자 하였다. 실험 결과 마지막 히든 레이어 임베딩을 활용하였을 때, 언어 모델의 성능은 66.4% 였다. 히든 레이어 별 언어 모델 성능을 비교한 결과, 마지막 4개의 히든 레이어를 이었을 때(concatenated), 언어 모델의 성능은 67.9% 이였으며, 11번째 히든 레이어를 사용했을 때는 68.1% 이였다. 즉, 마지막 히든 레이어를 선택했을 때보다 더 성능이 좋았다는 것을 알 수 있었다. 하지만 각 언어 모델 별 히트맵을 그려보았을 때는 마지막 히든 레이어 임베딩을 활용한 언어 모델이 더 정확히 의미역 판단을 한다는 것을 알 수 있었다.

• Journal of Korean Society of Steel Construction
• /
• v.23 no.4
• /
• pp.475-481
• /
• 2011

In this study, a simplified method of improving not only transverse shear stress but also shear strain based on the first-order shear deformation theory was developed. Unlike many established methods, such as the higher-order shear deformation and layerwise theories, this method can easily apply to finite elements as only $C^0$ continuity is necessary and the formulation of equations is very simple. The basic concept in this method, however, must be corrected:the distribution of the transverse shear stresses and shear strains through the thickness from the formulation based on the higher-order shear deformation theory. Therefore, the shear correction factors are no longer required, based on the first-order shear deformation theory. Numerical analyses were conducted to verify the validity of the proposed formulations. The solutions based on the simplified method were in very good agreement with the results considering the higher-order shear deformation theory.